sassine/sdd-desafio

Purpose and Delivery Model

This is a Spec-Driven Development challenge repository, not an implemented application. Participants fork it and produce a documented, traceable reimbursement-engine project. The central deliverable is the specification and its evidence trail; working code is secondary.

Core entry documents:

• DESAFIO.md defines the exercise and reimbursement-policy problem.

• RUBRICA.md defines evaluation criteria.

• README.md provides the workflow and required final repository shape.

Input and Domain Boundary

The reimbursement policy is intentionally ambiguous. The participant must derive explicit business rules, decisions, and acceptance criteria rather than seek clarification.

Reference input is:

• exemplos/despesas-exemplo.json

This JSON is the canonical sample expense dataset to inspect before designing rules, validations, classifications, reimbursement calculations, and expected outcomes.

Specification-First Workflow

The repository supplies a reusable project skeleton under:

• template/specs/001-motor-reembolso/spec.md

• template/specs/001-motor-reembolso/plan.md

• template/specs/001-motor-reembolso/tasks.md

• template/specs/001-motor-reembolso/DECISIONS.md

The intended flow is: interpret policy and sample data → write the product specification → select implementation design → decompose into numbered tasks with acceptance criteria → implement and test only against those tasks.

Decision and Change Control

Ambiguities are a first-class architectural input. Decisions must be explicit, justified, and retained in:

• template/specs/001-motor-reembolso/DECISIONS.md

• template/specs/001-motor-reembolso/spec.md

A mandatory Day-2 requirement change tests whether the specification can absorb changed policy without losing traceability. The decision log is the boundary between original requirements, interpretations, and later revisions.

Traceability and Execution Governance

Commits are governed by task IDs: no implementation commit should exist without a corresponding task. The required chain is specification → task → commit → test/result.

Agent collaboration conventions are seeded in:

• template/CLAUDE.md

Claude Code conversations must be exported and committed under the eventual docs/sessions/ directory, making the reasoning process auditable alongside code and documents.

Expected Generated Project

After copying the template, the participant’s fork should add root-level CLAUDE.md, specs/, and docs/, then create src/ and tests/. The final project is expected to expose a reimbursement engine, though this starter repository contains no runtime, source code, test framework, database, web service, or deployment infrastructure.

The final report is based on:

• template/docs/RELATORIO.md

It should connect the implemented outcome, specification decisions, changes, tests, commits, and exported sessions.